Rizal, Azri Agus and Ambia, Fajril and Rita, Novia and Herawati, Ira (2025) Studi Komparatif Capacitance Resistance Model dan Machine Learning Untuk Analisis Sensitivitas Kinerja Injeksi Polimer. Mechatronics, Electrical Power, and Vehicular Technology.

Text skripsi_213210089_watermark.pdf - Published Version Restricted to Registered users only Download (1MB) | Request a copy

Abstract

objective of this study was to evaluate the performance of polymer injection in the Volve Field by validating full-physics tNavigator simulation results. This process was performed using two independent data-driven approaches: the Capacitance Resistance Model (CRM) and machine-learning algorithms Random Forest and XGBoost. This validation framework addresses uncertainty in flow-parameter and ensures that simulated production responses align with data-driven injection–production behavior. The simulation model was constructed using 20 years of historical field data, consisted of five years of polymer injection at 1000–3000 ppm, followed by 15 years of chase water flooding. The simulation results showed that polymer injection increased the oil recovery factor from 21.12% to 21.30% in the best-case scenario, indicating a modest improvement in sweep efficiency. CRM, applied through CRMP and CRM-IP configurations, successfully reconstructed production profiles and quantified interwell connectivity (R² = 0.94; MAPE < 10%). Machine-learning validation further confirmed these results, with Random Forest achieving R² = 0.92 (MAPE < 1%) and XGBoost achieving R² = 0.99 (MAPE < 1%). Overall, CRM and machine learning provide effective and independent validation pathways, enhancing confidence in simulation outcomes and allowing for reliable assessment of polymer-injection performance in field applications.

Actions (login required)

View Item